Multireader diagnostic performance of MRI-based Node-RADS for pelvic lymph node metastasis in endometrial carcinoma.

Abstract

To evaluate the efficacy of the MRI-based Node Reporting and Data System (MRI-Node-RADS) in diagnosing pelvic lymph node metastasis (PLNM) in patients with endometrial carcinoma (EC). EC patients were retrospectively enrolled from July 2017 to August 2024. Two readers evaluated pelvic lymph nodes (PLNs) using MRI-Node-RADS. Pathological results served as the gold standard for determining the diagnostic accuracy of the scores. The criteria across size-based subregions were compared, focusing on obturator lymph nodes (Ob LNs) and non-obturator lymph nodes (non-Ob LNs). Inter-reader agreement was assessed using the weighted kappa statistic (κ_w). The area under the curve (AUC) was calculated to assess the sensitivity and specificity of the MRI-Node-RADS scores. Four hundred seventy-five PLNs were evaluated in 174 EC patients, comprising 85 metastatic and 390 non-metastatic PLNs. The inter-reader agreement was near-perfect at both evaluation levels: patient-level analyses (κ_w = 0.87) and regional analyses across eight pelvic locations (κ_w = 0.94). An MRI-Node-RADS score of  > 2 demonstrated optimal diagnostic performance, with an AUC of 0.93 (91.1% sensitivity, 84.5% specificity) at the patient level and 0.91 (85.9% sensitivity, 91.8% specificity) when analyzing individual PLN regions. The best performance among individual criteria was "Any change in texture" in Ob LNs and "Border: irregular or ill-defined" in non-Ob LNs. The MRI-Node-RADS effectively diagnoses PLNM, and a score of > 2 may be recommended as the optimal reference value for diagnosing PLNM in EC patients. Question Accurate assessment of PLNM is crucial for patients with EC, yet standardized guidelines for radiological reports are lacking. Findings An MRI-Node-RADS score of > 2 is identified as the optimal cut-off for diagnosing PLNM, with nearly perfect inter-reader agreement. Clinical relevance MRI-Node-RADS demonstrates excellent performance in diagnosing PLN metastases in patients with EC, suggesting that Node-RADS could be used as a reliable tool for clinical staging and personalized therapeutic decision-making.